Griptape Load: Audio

If you want Griptape to transcribe a recording or read text out loud, the pipeline starts here. Griptape Load: Audio takes a file sitting in your ComfyUI input folder and turns it into something the audio tasks can chew on.

This node exists because Griptape's audio tasks (transcription, text-to-speech) don't take a raw file path the way you'd hope - they expect an audio payload, or a path to one. This loader gives you both, which is why it's the standard entry point for any Griptape audio workflow.

How it works

When you open the node, its audio input is a dropdown listing audio files from ComfyUI's input directory. Drag a .wav, .mp3, .ogg, .flac, or .aiff file into ComfyUI, and it lands in that folder and shows up in the list. Pick it, and the node reads it with torchaudio, loading the waveform and sample rate, then hands you:

• AUDIO_PATH (string) - the absolute path to the file. Handy if the next step just needs to know where the file lives.
• AUDIO - the decoded waveform tensor plus its sample rate, ready for Griptape's audio transcription or TTS tasks.

It's a pure loader: nothing leaves your machine, no API key involved. The work happens in the nodes downstream.

Where it fits

The typical flow is Load: Audio → Griptape Run: Audio Transcription Task, with an agent that has an audio transcription driver configured. Or you pair it with a text-to-speech task to regenerate or summarize the content. If your audio doesn't appear in the dropdown, it's almost always a location problem - the file needs to be in the input folder (drag it onto the ComfyUI window and it lands there), not just anywhere on disk.

Installing

Part of the ComfyUI Griptape Nodes pack:

• ComfyUI Manager: search "Griptape" → install ComfyUI-Griptape.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/griptape-ai/ComfyUI-Griptape, then restart.

The pack brings in griptape[all], openai, python-dotenv, and git-hosted driver extensions via requirements.txt. Note the transcription side may pull heavier audio deps through griptape[all]. Same torch caveat as every Griptape install: Griptape declares torch and can fight ComfyUI's version - the README's fix is to uninstall and reinstall with --extra-index-url https://download.pytorch.org/whl/cu121.

Gotchas

The dropdown is built at node creation, so a file you just added may not appear until you refresh or recreate the node. Long filenames and weird characters in the path can also trip up the file lookup - keep it simple. And remember the output is a decoded tensor, not a reference you can hand to a normal ComfyUI audio node; it speaks Griptape's AUDIO type, so keep it inside the Griptape portion of your graph.